[OpenAFS-devel] release-team meeting minutes 2014-04-30

[Stephan Wiesand]

Log: =
http://conference.openafs.org/release-team@conference.openafs.org/2014-04-=
30.html

[Note the .html suffix instead of the former .txt one. This was the =
first meeting on the new infrastructure server hosted by MIT - thanks to =
Daria, Ben and Simon for the migration effort]

Participants:

* Andrew Deason
* Ben Kaduk
* Daria Brashear
* Jeffrey Altman
* Mike Meffie
* Stephan Wiesand

=3D=3D Next stable branch(es) =3D=3D

The RXGK changes queued in gerrit for master ( =
http://gerrit.openafs.org/#q,status:open+project:openafs+branch:master+top=
ic:rxgk,n,z ) have reached the point where people may want to start =
testing them, but merging them requires the new to-become-stable =
branches for Windows and Unix/Linux to be created before. Ben thinks =
that maintaining the current stack is not a huge burden right now, but =
the next step may be harder to maintain.

Alas, no progress. This topic obviously must be discussed and decided =
elsewhere.

=3D=3D Problem reports =3D=3D

=3D=3D=3D Fileserver crash in libkrb5 (RT #131852) =3D=3D=3D

Believed to affect Solaris and AIX, not Linux or FreeBSD, and only sites =
using Heimdal (which isn't trivial to build on the affected platforms). =
Affects all releases since 1.6.5. A simple workaround (touch =
rxkad.keytab) exists for all known occurrences of the problem but may =
cause others to go unnoticed. Gerrit 11075 works around the known =
instances of the problem without that side effect, but is not a complete =
solution either.

Decision: Andrew will bring up the issue on openafs-info, thus warning =
the (probably very few) sites that could encounter this issue. 11075 =
will be included in a future (pre-)release if it's still necessary.

=3D=3D=3D Linux callback processing failures with 1.6.8pre2 (RT #131855) =
=3D=3D=3D

Hard to investigate, because it only happens in the reporting site's =
production environment. Even test systems there with identical software =
setup don't exhibit the bug. Signature: when a directory is renamed and =
a new one created with the same name, other clients may see the old =
content. More information will be added to the RT entry. Reverting 10948 =
and 10804 fixes the problem but makes the getcwd ENOENT problem =
reappear.

Decision: We'll wait for another week for results of the ongoing =
investigation. If this can't be sorted out in time, we may revert the =
two changes before the 1.6.8 release. If there's no other change, this =
may not even warrant a pre3.

=3D=3D Linux news =3D=3D

3.15-rc3 was released. Marc tested against rc3+ yesterday, and we're =
still good.

Marc also reports that most likely patches will be required for 3.16, =
which is expected to be released in ~4.5 months. This may make it =
feasible to address 3.16 in 1.6.9 rather than a 1.6.8.1.

=3D=3D 1.6.8 release =3D=3D

Stephan uploaded the last pre2 binaries for RHEL/Fedora built by Stephen =
in Edinburgh today. Despite the known problems described above, testing =
pre2 still makes sense since sites are unlikely to hit them [and if they =
do, all the better]. The new binaries will be announced on -info.

=3D=3D Testing =3D=3D

Now that the nature and scope of the known problems is more clear, =
Andrew will approach the sites that volunteered for testing prereleases =
about 1.6.8pre2.

--=20
Stephan Wiesand
DESY -DV-
Platanenenallee 6
15738 Zeuthen, Germany